Cost-effectiveness of Total Neoadjuvant Therapy With Short-Course Radiotherapy for Resectable Locally Advanced Rectal Cancer

Key Points Question For locally advanced rectal cancer, what is the economic implication of short-course radiotherapy and total neoadjuvant therapy followed by total mesorectal excision compared with conventional long-course chemoradiotherapy followed by total mesorectal excision? Findings This decision analytical model found that compared with conventional long-course chemoradiotherapy with or without adjuvant chemotherapy, short-course radiotherapy and total neoadjuvant therapy were associated with cost savings without diminishment in quality-adjusted life-years. Meaning The findings of this study support short-course radiotherapy and total neoadjuvant therapy as a new treatment paradigm in the management of locally advanced rectal cancer.


Introduction
Colorectal cancer is the second leading cause of cancer-related mortality in the US, with the country's second-highest annual cost of $14.1 billion in 2010. 1 Costs were projected to reach $17.4 billion in 2020. 1 Standard of care in the management of locally advanced rectal cancer usually entails neoadjuvant long-course chemoradiotherapy (LCCRT) for 5 to 6 weeks, 2,3 followed by total mesorectal excision (TME). 4,5 Although this treatment strategy has led to decreased local recurrence rates of 4% to 9%, [5][6][7] distant metastases remain the predominant site of recurrence, 8 and the management of metastatic rectal cancer incurs significant cost and morbidity. 9 To improve tumor downstaging before surgery, decrease the rates of distant metastases, and improve chemotherapy adherence, investigators more recently adopted a total neoadjuvant therapy (TNT) approach before TME. [10][11][12] Adding multiagent chemotherapy to the interval between radiotherapy and surgery has been shown to improve tumor downstaging 13 and chemotherapy tolerance. 14,15 Notably, in the phase 3 international multicenter trial Rectal Cancer and Preoperative Induction Therapy Followed by Dedicated Operation (RAPIDO), preoperative short-course radiotherapy followed by TNT (SCRT-TNT) led to an increased pathological compete response rate, decreased disease-related treatment failure, and decreased distant metastatic disease at 3 years compared with preoperative LCCRT with or without adjuvant chemotherapy. 10,16 Although SCRT-TNT has shown oncologic promise and is recommended by the National Comprehensive Cancer Network, 17 the economic impact of this new therapy is not fully understood.
Previous cost-effectiveness analyses of treatment paradigms for locally advanced rectal cancer have compared conventional LCCRT with SCRT alone, 18 SCRT with a short duration of consolidation chemotherapy, 19 and long-course TNT. 20 However, there are no economic evaluations comparing conventional LCCRT with SCRT-TNT. Therefore, we performed a cost-effectiveness analysis of SCRT-TNT vs conventional LCCRT using data from the RAPIDO trial 10,16 and other published data.

Decision Analytical Model
A decision analytical Markov model with a 5-year time horizon was designed to compare SCRT-TNT vs LCCRT followed by TME for patients with locally advanced (T3-T4 or node-positive) adenocarcinoma of the rectum using TreeAge Pro software, version 2020 R2.1 (TreeAge Software, LLC) ( Figure 1A). This time horizon was selected because the oncologic outcomes were assumed to be the same between the 2 treatment groups 5 years after treatment completion.
We defined SCRT as 25 Gy in 5 fractions for 5 treatment days with 3-dimensional techniques. 5,6,[22][23][24] We defined SCRT-TNT as neoadjuvant SCRT followed by consolidation chemotherapy with the modified FOLFOX regimen (leucovorin calcium [folinic acid], fluorouracil, and oxaliplatin) for 9 cycles 13,25 or the CAPOX regimen (capecitabine and oxaliplatin) for 6 cycles. 15,26,27 We defined LCCRT as 50.4 Gy in 28 fractions for 28 treatment days with concurrent capecitabine. 28 Subsequently, TME was performed with either abdominoperineal resection and a permanent colostomy 29 or low anterior resection with a temporary defunctioning ileostomy and planned reversal. 30,31 The base model included adjuvant chemotherapy after LCCRT per the National Comprehensive Cancer Network guideline, 32 although the delivery of such treatment was at the discretion of the treating hospitals in the RAPIDO trial. 10,16 After treatment with either SCRT-TNT or LCCRT, Markov models were constructed to describe disease progression and patient survival.
The Markov model as illustrated by the transition state diagram was characterized by 5 health states: no evidence of disease (NED), resectable locoregional recurrence (LRR), unresectable LRR, distant metastatic disease, and all-cause death (absorbing state) ( Figure 1B). Furthermore, a temporary health state of LRR was constructed to reflect the differing costs and utilities associated with resectable and unresectable LRR. The cycle length of this Markov model was 3 months, which was chosen to model the real-world intervals between office visits, staging imaging, and subsequent treatment decisions.
After TME, all patients entered the NED state of the Markov model. In the next cycle, they could remain in this state or transition to LRR, distant metastatic disease, or death. All patients with LRR were assumed to have received a second course of radiotherapy using the previously established hyperfractionated accelerated regimen of 39 Gy in 26 fractions twice a day delivered for 13 treatment days. 33 A proportion of patients with LRR were assumed to have resectable disease and underwent additional salvage abdominoperineal resection or pelvic exenteration. 34 For patients with unresectable LRR or distant metastatic disease, the patients were assumed to have received palliative chemotherapy with capecitabine for 1 year. For patients with unresectable LRR, it was assumed that the patients could stay in the unresectable LRR state or transition to death.

Probabilities
The probabilities of undergoing abdominoperineal resection and low anterior resection after SCRT-TNT or LCCRT were derived from the results of the RAPIDO trial 10,16 ( Table 1). The risk of progression between states was governed by the transition probabilities in the Markov model and differed by treatment strategies. The probabilities of NED to LRR and NED to distant metastatic disease after SCRT-TNT and LCCRT in the RAPIDO trial 10,16 were used for transition probabilities in the first cycle (Table 1). Beyond the first cycle, we assumed that the probability of transitioning from any state (resectable or unresectable LRR to distant metastatic disease or death) was the same for each treatment group 35 (Table 1).  amounts were used to estimate the admission cost associated with TME. 40 An annual ostomy maintenance cost was assumed for patients who underwent initial abdominoperineal resection. All costs were adjusted to 2020 US dollars using the consumer price index. 41,42

Cost-effectiveness Analysis
Markov cohort analysis with half-cycle correction was performed to compute the total health care costs and QALYs after each treatment accumulated during the 5-year time horizon. Cost, QALYs, and utilities were discounted at an annual rate of 3%. 41 For each treatment strategy, the 3-year LRR, cumulative distant metastasis, and overall survival rates were computed from the model.
The incremental cost-effectiveness ratio (ICER), defined as the ratio of the incremental cost and the incremental QALY gained, was calculated to compare the cost-effectiveness of these treatment paradigms. The net monetary benefit was defined as the QALYs multiplied by the willingness to pay (WTP) per QALY gained subtracted by the total cost. The WTP threshold was defined as $50 000/ QALY for strategies that were clearly cost-effective in the base case. 43

Sensitivity Analysis
Sensitivity analyses were conducted to test the robustness of the conclusion. In addition, WTP threshold was varied to $100 000/QALY and $150 000/QALY. 43 Multiple 1-way sensitivity analyses were performed for the probability, utility, and cost parameters derived from Table 1 and Table 2 with the variable range set to plus or minus 25% of the base case values and presented through a tornado diagram (Figure 2). Variables that had significant variability in published values (ie, utility of NED after abdominoperineal resection vs NED after low anterior resection) 36,44 or had the greatest potential for fluctuations over time with improvement in technology or policy (ie, cost of SCRT-TNT and LCCRT) were evaluated in 2-way sensitivity analyses. Two-way sensitivity analyses were performed by varying the influential variables determined in the 1-way sensitivity analyses. Data were analyzed from November 15, 2020, to April 25, 2021.   for LCCRT. We also repeated the analysis assuming (1) adjuvant chemotherapy was given to all patients in the LCCRT cohort and (2) the rates of locoregional and distant recurrences disease were equal between the SCRT-TNT and LCCRT cohorts (ie, assumed to be 8% and 25%, respectively, at 3 years). Short-course radiotherapy followed by TNT still incurred a lower total cost with similar QALYs in this hypothetical scenario. The total cost was $41 380 and the QALYs were 2.14 for SCRT-TNT, and the total cost was $55 399 and the QALYs were 2.13 for LCCRT during the 5-year horizon. This resulted in a negative ICER.

1-Way Sensitivity Analysis
The tornado diagram for the multiple 1-way sensitivity analyses is shown in Figure 2. The most influential variables affecting model robustness were the probabilities of transitioning from NED to distant metastasis for SCRT-TNT and LCCRT, the probabilities of low anterior resection after SCRT-TNT and LCCRT, the utility of being in NED after low anterior resection, and the cost of LCCRT and   adjuvant chemotherapy. In all instances, differing each variable by 25% around the base values resulted in ICERs that remained consistent with the base case, which illustrated that SCRT-TNT was the preferred cost-saving strategy over LCCRT. Furthermore, because adjuvant chemotherapy was given at the discretion of the treating hospital in the RAPIDO trial, 10,16 the cost of adjuvant chemotherapy was also decreased from the base case of $6885 to zero in the sensitivity analysis to reflect either complete or no adjuvant chemotherapy use. Short-course radiotherapy followed by TNT remained the cost-saving strategy after eliminating the cost of adjuvant chemotherapy. The conclusions of the 1-way sensitivity analyses were upheld at a WTP of $100 000/QALY and $150 000/QALY (Figure 2). At a WTP threshold of $50 000, SCRT-TNT remained the preferred strategy unless the cost of SCRT-TNT exceeded $27 607, which was 2.8 times the cost of SCRT-TNT assumed in the base case.

2-Way Sensitivity Analysis
We

Discussion
Short-course radiotherapy followed by TNT has emerged as a potential treatment paradigm in the management of locally advanced rectal cancer. Despite the emerging evidence for SCRT-TNT, data comparing the cost-effectiveness of SCRT-TNT with conventional LCCRT are scarce. This study uniquely demonstrates the cost-saving economic advantage of SCRT-TNT compared with LCCRT with or without adjuvant chemotherapy using data from a single prospective phase 3 randomized clinical trial.
Our results corroborate the analyses from a previously published economic study by Raldow et al, 18 which demonstrated that LCCRT was not cost-effective compared with SCRT with an ICER of $133 495/QALY when combined with conventional adjuvant chemotherapy using data from the German rectal trial. 2 These results also are consistent with those of Wang et al, 19 which demonstrated that SCRT with consolidation chemotherapy was more cost-effective than LCCRT with or without adjuvant chemotherapy using data from the Polish II trial, 25,45 from the perspective of a Chinese payer. Notably, the consolidation chemotherapy regimen in the SCRT group of the Polish II trial only used 3 cycles of FOLFOX4 (FOLFOX regimen including both a bolus and infusion of fluorouracil), 25,45 whereas the RAPIDO trial used CAPOX for 6 cycles or FOLFOX4 for 9 cycles. 10, 16 Wright et al 20 also showed that long-course TNT was cost-effective compared to LCCRT with adjuvant chemotherapy.
At present, SCRT is underused (<1%) in the US, 46 but it is gaining traction in the setting of increased interest in shortening treatment in the setting of the COVID-19 pandemic. 47 Previous studies have suggested that SCRT might be less efficacious with less tumor downstaging compared with LCCRT 48 and might result in more acute toxic effects. 49 However, the Stockholm III trial 50,51 showed greater tumor downstaging and decreased postoperative complications in patients treated with SCRT and delayed surgery compared with long-course radiotherapy (without concurrent chemotherapy) and delayed surgery. Delaying surgery after SCRT also decreased the rate of postoperative complications compared with immediate surgery. 50 The phase 3 RAPIDO trial 10,16 recently showed that SCRT followed by consolidation chemotherapy and TME increased the rate of

Strengths and Limitations
To our knowledge, this decision analytical model is the first reported economic evaluation of SCRT-TNT and LCCRT. We performed detailed time-dependent modeling of health states using randomized clinical trial data published in the modern era and included a comprehensive microcosting analysis. The results of this study support future exploration of SCRT-TNT in the management of locally advanced rectal cancer. Adoption of this treatment paradigm should also await quality of life and patient-reported outcomes data as well as maturing, long-term oncologic survival data.
This study also has multiple limitations. The model compared conventional LCCRT with a novel and emerging TNT-based regimen, which was only recently reported in phase 3 trials 10,61 and is still being assessed in ongoing trials for the management of locally advanced rectal cancer. 53,60 No quality of life or patient-reported outcome measures from these TNT studies have been published to date.
The definition of locally advanced rectal cancer of cT3 to cT4 or node-positive disease in this study encompasses more patients than those enrolled in the RAPIDO trial with high-risk factors such as cT4a, cT4b, or cN2 disease, extramural vascular invasion, involved mesorectal fascia (tumor or lymph node Յ1 mm from the mesorectal fascia), or enlarged lateral lymph nodes considered to be metastatic. 10,16 In terms of data sources, the tumor recurrence and survival outcomes were primarily based on a single phase 3 study (RAPIDO), and the cost was entirely derived from US-based Medicare data. Although these factors could potentially limit the broader applicability and generalizability of the study, the conclusions were upheld in the sensitivity analyses that were performed.
As for the model structure, the Markov model relied on simplification of disease processes and costs and was limited by the quality of data used to generate probabilities, utilities, and costs. The use of adjuvant chemotherapy in the LCCRT group was left to the discretion of the participating hospitals in the RAPIDO trial, which resulted in some treatment heterogeneity, although subgroup analysis demonstrated similar oncologic outcomes between these 2 groups. 10,16 The conclusion that SCRT-TNT was cost-saving was upheld whether compared with LCCRT with or without adjuvant chemotherapy. In the extreme scenario wherein the locoregional and distant recurrence rates were assumed to be the same and all patients were assumed to have received adjuvant chemotherapy, SCRT-TNT was found to result in an even greater magnitude of cost saving while achieving a similar amount of QALYs as LCCRT.

Conclusions
The findings of this decision analytical model suggest that SCRT followed by TNT and TME was associated with superior oncologic outcomes and lower cost compared with conventional LCCRT followed by TME with or without adjuvant chemotherapy. In the context of large randomized clinical trial data demonstrating superiority of SCRT-TNT to LCCRT, the presented data support the exploration of SCRT-TNT as a new cost-saving treatment paradigm in the management of locally advanced rectal cancer.